Catalyst deactivation during the hydrogenation of benzene over nickel-loaded Y zeolites

Catalyst deactivation during the hydrogenation of benzene over nickel-loaded Y zeolites

Jounml of Molecukw Catalysis, 71 (1992) 93-109 MOLCAT 2783 93 Catalyst deactivation during the hydrogenation of benzene over nickel-loaded Y zeolite...

1MB Sizes 0 Downloads 8 Views

Jounml of Molecukw Catalysis, 71 (1992) 93-109 MOLCAT 2783

93

Catalyst deactivation during the hydrogenation of benzene over nickel-loaded Y zeolites Brendan Coughlan and Mark A. Keane* Physical Chemistry L&oratories,

University College, Galway (7?-eiandj

(Received May 2, 1991; revised September 10, 1991)

Abstract The hydrogenation of benzene to cyclohexane was investigated over a range of nickelexchanged and nickel-impregnated Y zeolites, varying the nickel content and the nature of the alkali metal co-cation (Li+, Na+, K+, Rb+ or Cs’). With a view to optimizing benzene conversion levels, the following catalytic parameters were studied: reaction temperature, reaction time, benzene flow rate and coke deposition. The observed catalytic activities are correlated with previously reported physical characterizations. Benzene hydrogenation increased in the order: NiLiY < NiNaY < NiKY < NiRbNaY < NiCsNaY. Catalyst deactivation results from the deposition of involatile coke on the catalyst surface, which is promoted by increasing zeolite acidity. The effects of poisoning the surface Brijnsted acid sites by adsorption of ammonia onto the activated reduced zeolites are considered. The results of catalyst regeneration by high temperature oxidation of the coke deposits are also reported.

Introduction

Metal catalyst deactivation or aging is caused by any process which obliterates the active sites or blocks access to them by the reactant molecules. In general, any or a combination of the following factors may be involved: (i) metal poisoning due to the presence of sulphur in the feedstock; (ii) sintering, which leads to a loss of metal surface area; (iii) coking. The interaction between sulphur (or sulphur-containing compounds) and metallic surfaces has been the subjected of two detailed reviews [l, 21. Sintering of supported metal particles to form large metal crystallites is well recognized [3-S]. Catalyst deactivation by coking has received a comprehensive treatment in three recent books [ 7-91. Carbonaceous residues are the inevitable byproducts of most heterogeneously catalyzed organic conversions [ 8 1.The term ‘coke’ designates such deposits which often encompass a range of heavy polynuclear aromatic molecules. The loss of catalytic activity caused by the formation of coke in zeolites results from a poisoning of the intracrystalline active sites and a plugging of the zeolite pores. Zeolite coking *Author to whom correspondence should be addressed at: Chemistry Department, The University, Glasgow G12 8&Q, UK.

0304-5102/92/$5.00

0 1992 - Elsevier Sequoia. AU rights reserved

94

is therefore spatially demanding and essentially analogous to a reverse molecular shape selectivity. Since the pioneering work of Voorhies [lo], there have been many attempts to model the deactivation process (11-131. Voorhies [lo], in his ‘time on stream’ theory, proposed that the percentage carbon deposited on the catalyst (C,) followed the relation C,=AV, where t is the process time and A and n are constants. This approach, though still in use [ 11, 121, suffers from the disadvantage that each relationship is specific to a single reaction and should only be applied under the reaction conditions for which it is formulated. Froment and Bishoff [ 131 have argued that if catalyst deactivation is due to coke deposition and coke formation depends on the concentration of reacting species, then catalyst deactivation cannot be a simple function of time. Rather, they have suggested a model in which the rate of coke formation is treated in the same manner as the main reaction with the result that the deactivation function is related to the coke content as opposed to the ‘time on stream’. Although the process of deactivation by coking has been well documented, the methods used for regenerating coked catalysts have received little attention. The normal industrial and laboratory procedure is to simply oxidize the coke to carbon oxides and water by heating the deactivated catalysts in oxygen in the temperature range 673-773 K 114-161. Copperthwaite et al. [ 171 have also proposed a lower temperature oxidation using oxygen/ozone mixtures. Recent studies conducted in this laboratory [ 18, 19 ] have been devoted to the kinetics and mechanism of benzene hydrogenation over nickel-exchanged Y zeolites. The present report is concerned with probing the catalytic activities of a range of nickel Y zeolites enriched to various levels with different alkali metal co-cations. In particular, the extent of catalyst deactivation is studied and the possibilities for spent catalyst reactivation are considered.

Experimental

The starting or parent zeolite was Linde Molecular Sieve LZY-52 (formula: NaSs(A10,),,(Si0,)134(H20)280). The ion-exchange and impregnation procedures used to prepare the nickel-loaded zeolites are detailed elsewhere [6, 20 ]. All samples were stored over saturated NH&l solutions in order to maintain constant humidity. The ion-exchanged samples are labelled according to the % exchange of indigenous alkali metal co-cation and the total number of Ni2+ ions * per unit cell, e.g. the NiKY-23.5 sample exhibits a 23.5% exchange of the 58 parent K+ ions, resulting in a total of 6.8 Ni2+AJ.C., whereas the impregnated samples are labelled according to percentage by weight of nickel, e.g. in the case of the NiO-KY/2.2 sample, 2.2% of the total weight is nickel which equals 6.0 Ni2+ /U.C.. Nickel cry&allite size measurement was carried out by X-ray diffraction line broadening (Jeol JDX-85 Diffractometer) around the nickel line at a 28 angle of 52.2” under the conditions described previously

95

[ 61. X-ray diffraction and infrared spectroscopy were used to monitor the crystallinity of the samples [6,20]; the IR band at 395 cm-’ has been shown to be very sensitive to changes in zeolite crystallinity [ 2 11. The catalytic reactor and procedure used in this study have been described in detail in a previous paper [ 181. Chromatographic analyses were carried out on the liquid samples, collected in a liquid nitrogen trap, using a 5% Bentone/5% di-isodecylphthalate on Chromosorb W chromatography column with a carrier gas (N,) flow rate of 20 cm3 min-’ and column temperature at 348 K. The effluent from the reactor was collected over half-hourly periods and the percentage conversion after this time interval was equated to the percentage cyclohexane in the sample. Mel% conversions after 6 h on stream were used as a point of comparison between different catalysts. The W/F values quoted in this study represent the mass of hydrated catalyst in grams divided by the feed rate of reactants in moles per hour. Treatment of the reduced zeolite samples with ammonia to poison the surface Briinsted sites has been described in detail elsewhere [ 61. The extent of occluded ammonia was monitored by measuring the nitrogen content of the zeolite samples by microanalysis. Similarly, the levels of coke formation were inferred from carbon microanalysis of the spent zeolites. Oxidative regeneration of the deactivated samples was carried out by heating the zeolite pellets in situ in the catalytic reactor [18] at 50 K h-’ in a 120 cm3 min-’ stream of oxygen to a final temperature of 723 K which was maintained for 1 h. The oxygen flow was then replaced by hydrogen flow and the samples were reduced for 18 h at 723 K. The temperature was then lowered to 473 K prior to catalysis. Thermogravimetric analyses (Perkin Elmer Thermobalance) were conducted on the spent samples in order to measure the weight loss as a function of time during the regeneration process; the samples were heated in a 20 cm3 min-’ stream of oxygen at 40 K min- ’ in the temperature range 294-773 K. The benzene used in this study was of AnalaR grade and was further dried by standard methods [ 22 ].

Results

and discussion

The chemical compositions of the ion-exchanged and impregnated samples are tabulated elsewhere [ 61. X-ray diffraction and IR analyses showed that crystallinity was maintained for every sample after catalysis. Effect of reaction time For a complete understanding of the reaction system, it is necessary to study the shape of the activity profiles as a function of reaction time. The hydrogenation of benzene over all the nickel-loaded zeolites generated activity VS. time prollles characterized by an initial drop in activity over the first hour on stream, which is followed by either attainment of a steady state conversion or by a continuous slow deactivation which eventually rendered the catalysts totaI.ly inactive, Fig. 1. Only those samples which were acidic

96

Time onstream

I hr

Fig. 1. Variation of benzene hydrogenation with reaction time over ion-exchanged NUY-43.1 (12.4 Ni’+/U.C.) (X), NiKY-23.5 (6.8 Ni’+/U.C.) (O), NiNHIY-52.2 (15.1 Ni’+/U.C.) (A) and impregnated NiO-KY/l.9 (6.0 Ni2’/U.C.) (W) samples; T=473 K, W/F=134.1 g mole1 h.

in nature, either through the introduction of acidic hydroxyls by the exchange of the parent alkali metal ion with ammonium ions followed by thermal deammoniation (as exemplified by the NiNH4Y-52.2, 15.1 Ni2+/U.C. sample in Fig. 1) or by reduction of the higher nickel-loaded samples [23] (as in the case of NiLiY-43.1, 12.4 Ni2+AJ.C.), exhibit this continuous deactivation. The rate of deactivation therefore increases with increasing sample acidity. This is consistent with the previously reported mechanistic study carried out in this laboratory on the same catalytic system [ 191. The decrease in the proportion of cyclohexane in the product mixture over the first hour on stream is a feature of all the activity profiles, Fig. 1. In addition, the volume of the first two liquid samples (the first in particular) collected from the reactor was much smaller than the subsequent samples. This suggests that a sorption phenomenon is in effect under the reaction conditions. From IR spectroscopic and chromatographic measurements [24, 251, it has been observed that benzene interacts much more strongly than cyclohexane with the active zeolite surface under catalytic conditions. During the initial period of catalysis the zeolite surface selectively adsorbs benzene in preference to cyclohexane, which results in a relative enrichment of the first two samples with cyclohexane; this would also account for the relatively small volumes of the initial product samples. The observed conversion levels over the ilrst hour therefore do not represent the true activity of the catalyst. As the reaction proceeds, the sorption capacity of the zeolite surface for benzene becomes saturated and the selective adsorption phenomenon is no longer operational. The W/F parameter is a measure of the number of active sites available for the conversion of each molecule of reactant; the greater the value of W/F, the greater the probability that a reactant molecule will encounter an active site and be converted to product with an overall increase in reaction rate. The effect of lowering the value of W/F on the level of benzene hydrogenation over a range of nickel zeolites is shown in Table 1. It can be observed that a decrease in W/Flowers the extent of cyclohexane formation. This effect is not as evident for the higher nickel-loaded NaY and LiY samples,

97 TABLE

1

Effect of varying W/F on the level of benzene hydrogenation over a range of nickel-exchanged Y zeolites Zeolite sample

Ni”AJ.C.

Cyclohexane WIF=

NiLiY-8.8 NiLiY-43.1 NiLiY-63.7 NiNaY-6.8 NiNaY-35.7 NiNaY-48.8 NiNaY-63.1 NiKY-8.0 NiKY-30.5 NiKY-49.1 NiKY-62.5 NiRbNaY-7.3 NiRbNaY-35.1 NiRbNaY-59.1 NiCsNaY-3.6 NiCsNaY-3 1.1 NiCsNaY-54.8

“g mol-’

2.6 12.4 18.5 2.0 10.4 14.1 18.3 2.3 10.3 14.2 18.1 2.1 10.2 17.1 1.0 9.0 15.9

55.0 3.2 0.5 53.6 6.2 1.9 0.9 55.2 39.0 87.3 68.4 49.7 86.6 73.0 46.3 89.8 77.9

134.1a

(after 6 h on stream) at 473 K

(Mel%) W/F= 53.5

W/F= 26.8

20.2 0.8

8.9

21.4 1.9 0.6 0.1 22.2 12.3 37.6 26.3 21.6 30.4 18.3 19.4 31.2 26.1

10.4 0.7

10.9 3.9 10.8 5.2 10.4 12.2 6.1 9.6 14.4 9.4

h.

where the overwhelming influence of sample acidity rendered the catalysts inactive in any case. For all the samples studied, the initial drop in activity was not as pronounced at lower W/F values (increasing benzene flow) due to the shorter residence times of benzene on the zeolite surface at the higher organic flow rates. Eflect of nickel loading In order to characterize the catalytic activity of the supported nickel metal, the mol% conversion to cyclohexane after 6 h on stream is plotted vs. nickel loading for a range of Y zeolites in Fig. 2. In the case of the more acidic catalysts, which have been shown to undergo continuous deactivation, the levels of conversion after 6 h do not represent a steady state but rather a point by which comparisons may be made between different catalysts. Figure 2 illustrates a marked divergence in behaviour between the NiNaY, NiLiY, NiKY, NiRbNaY and NiCsNaY samples. Considering the NiNaY system, at nickel loadings greater than ca. 7 Ni2’/U.C. the extent of benzene hydrogenation decreases with increasing nickel content; the higher-loaded samples ( > 10 Ni” /IJ.C.) exhibit negligible activity after 6 h on stream and are completely deactivated after 8 h. In contrast, the NiKY system, while also exhibiting an initial decrease in activity with nickel exchange, is characterized by a remarkable increase in benzene conversion in the range 12-18 Ni2’/U.C.. This corresponds to a previously reported [25, 261 increase in

98

3 5 60. ..f > CI 40-t. > a, g 20.

+m +m l**+

+:

: 80. t.’

.-

..

.

.*.A=

. l

.

A .

l

.

6

.

’ .

x

.

. .

.

I x. . 12 18 N I 2+/ U.C.

l

24

Fig. 2. Variation of benzene hydrogenation (after 6 h on stream) as a function of nickel loading

over a range of ion-exchanged NiLiY (X), NiNaY (A), (+) samples; T=473 K, W/F=134.1 g mol-’ h.

NiKY (O), NiRbNaY (m) and NiCsNaY

Ni” reduction (and mass of supported nickel metal) over the same level of nickel exchange. The hydrogenation activity again drops at higher degrees of nickel exchange, and the nickel saturated samples (> cu. 18 Ni2+) eventually deactivate to zero benzene conversion after cu. 20 h on stream. The NiLiY samples exhibit behaviour similar to the NiNaYsystem, but catalyst deactivation becomes apparent at a much lower level of nickel exchange. By comparison, the NiRbNaY and NiCsNaY samples show a steady increase in activity with nickel loading, resulting in distinct maxima at cu. 13 Ni2+/U.C.. The highestexchanged catalysts (NiRbNaY-59.1: 17.1 Ni2+/tJ.C. and NiCsNaY-54.8: 15.9 Ni2+ /U.C.) show a slow rate of deactivation which results in a marked loss in activity (- 10% cyclohexane) after 40 h on stream. This behaviour is again entirely analogous to that observed in the degree of Ni2+ reduction in studies reported elsewhere [25, 261. The observed dependence of nickel cation reduction on the basicity of the charge-balancing alkali metal cocation has been explained by the influence of the latter in directing the siting of the Ni2+ cations in the anhydrous zeolite and influencing the location and strength of the surface Brijnsted acid sites [23]. In the absence of a high level of surface acidity, benzene hydrogenation is proportional to the extent of Ni2+ reduction and hence the mass of supported nickel metal. This concept was further developed by carrying out the reaction over a series of NiKNaY samples containing a constant amount of nickel (CCL14 Ni2+AJ.C.) but a varying K+ /Na+ ratio. Increasing the K+ /Na+ ratio, which has been shown to increase the level of Ni2+ reduction [ 261, resulted in a marked enhancement of hydrogenation activity, Table 2. In the same way, it has been observed that the level of reduction of Ni2+ cations supported on the NiKY-23.5 (6.8 Ni’+/U.C.) sample increased at lower hydrogen flow rates [20], with a resultant enhancement in hydrogenation activity as shown in Table 3. These observations lend further credence to a dependence of hydrogenation activity on the mass of nickel metal generated during the reduction process. Although cyclohexane was the major product, methylcyclopentane (MCPa) was also isolated in the product mixture. In our proposed mechanism [ 19 1,

99 TABLE 2 Mel% cyclohexane (after 6 h on stream) over a range of NiKNaY samples of varying K’/Na’ content: W/F=134.1 g mol-’ h, T=473 K Zeolite sample

Na+/U.C.

NiNaY-48.8 NiKNaY-47.7 NiKNaY-47.1 NiKNaY-49.3 NiKNaY-46.1 NiKNaY-48.4 NiKY-46.2

30.3 23.9 15.2 10.1 7.8 3.8

TABLE

K+ /U.C.

Ni2+/U .C .

Cyclohexane (mol%)

7.0 15.8 20.3 23.7 27.1 32.2

14.1 13.8 13.7 14.3 13.4 14.0 13.4

1.6 4.4 26.6 51.1 59.3 77.9 89.1

3

Effect of catalyst reducing conditions (hydrogen flow rate) on the level of benzene hydrogenation over NiKY-23.5 (6.8 Ni2+/CJ.C.) after 6 h on stream: W/F= 134.1 g mol-’ h; T=473 K Hydrogen flow rate (cm’ min-I)

Cyclohexane (mol%)

30 60 90 120 150 200 250 300 350 400

83.1 82.0 81.5 81.0 72.8 64.3 60.2 54.6 52.0 51.8

MCPa formation is viewed as resulting from the desorption of cyclohexene (as a reactive intermediate) from a metal site and its resultant migration to an acid site at which isomerization to methylcyclopentene (MCPe) is promoted; hydrogenation of MCPe to MCPa follows at a metal site. Considerable surface acidity is therefore necessary to effect MCPa formation. Moreover, the formation of MCPa decreased with time on stream and was promoted with increasing temperature and W/F value, Table 4. Mel% conversion to MCPa as a function of nickel loading is plotted in Pig. 3. It can be observed that % MCPa formation is small in comparison with cyclohexane production. Nevertheless, the extent of MCPa formation increases with nickel loading and hence with zeolite acidity in the order: NiCsNaY - NiRbNaY < NiKY < NiNaY < NiLiY; MCPa only appears in trace quantities after catalysis over the highest-loaded NiRbNaY and NiCsNaY samples. The dual functionality of the catalysts modifies the product mixture; the metallic activity of supported metal and the acidic activity of the surface

100

TABLE 4 Effect of varying W/F and/or reaction temperature on the level of methylcyclopentane (MCPa) formation (after 6 h on stream) during hydrogenation of benzene over a range of nickelexchanged Y zeolites Zeolite sample

Ni”/U.C.

MCPa (mol%) W/F= 134.1’ T=473

NiLiY-43.1 NiLiY-63.7 NiNaY-48.8 NiNaY-63.1 NiNaY-90.1 NiKY-62.5 NiKY-86.6

12.4 18.5 14.1 18.3 26.1 18.1 25.1

2.2 3.7 1.5 2.2 3.3 1.4 2.6

K

WIF= 53.5 T=473 K

WIF= 53.5 T=423 K

1.2 2.1 0.9 1.2 1.9 0.6 1.2

0.5

1.1 0.4 0.6 0.9 0.6

‘g mol-’ h.

Fig. 3. Variation of methylcyclopentane (MCPa) formation (after 6 h on stream) during hydrogenation of benzene as a function of nickel loading over a range of ion-exchanged NiLiY (X), NiNaY (A), NiKY (0) samples; T=473 K, W/F= 134.1 g mol-’ h.

Bronsted acid sites interact to yield a combined activity. As a result, the factors which establish the acidic properties of the support play a part in altering the hydrogenation activity of the metal function.

E8Jec.tof reaction temperature The reaction temperature for maximum hydrogenation activity has been established as 473 K for the non-acidic NiKY-23.5 (6.8 Ni” ./UC.) sample [ 181.This observed temperature of reaction rate maximum has been explained by a decrease in the surface coverage by benzene with increasing temperature, which at temperatures above 473 K is severe enough to lower the overall reaction rate 1181.However, the temperature-related maximum is shifted to lower temperatures for catalysis over the higher nickel-loaded, more acidic zeolites, as is clearly illustrated in Pig. 4; this effect is particularly marked in the case of the NiNaY samples. MCPa formation is inhibited at lower reaction temperatures. Under these conditions, the reactive cyclohexene

101

” Y-

32.

.

:

A

A! o 16.: Ix

u s

A 16.

*

A A

410

440

470 T/K

A .

* A

_&A* 500

;

4 Heating

6

8

RateiKhr’.lO

4. Effect of reaction temperature on level of benzene hydrogenation over NiKY-23.5 (6.8 Ni’+/U.C.) (O), NiKY-62.5 (18.1 Ni’+/U.C.) (O), NiNaY-22.7 (6.6 Ni2+/U.C.) (A) and NiNaY63.1 (18.3 Ni’+/U.C.) (A) after 6 h on stream; W/F=134.1 g mol-’ h. Fig.

Fig. 5. Effect of catalyst reducing conditions (rate of heating in hydrogen) on level of benzene hydrogenation over NiKY-23.5 (6.8 Ni”/U.C.) (a), NiKY-62.5 (18.1 Ni’+/U.C.) (O), NiNaY22.7 (6.6. Ni”/U.C.) (A) and NiNaY-63.1 (18.3 Ni’*/LLC.) (A) after 6 h on stream; T-473 K, W/F=134.1 g mol-’ h.

intermediate is more strongly held on the zeolite surface [ 19 1, with the result that MCPa formation is suppressedwhile cyclohexane formation is promoted. As previously discussed [ 201, an increased rate of heating of the nickel zeolites in hydrogen, during the reduction process, lowers the contact time of hydrogen with the partially dehydrated Ni2+ ions which migrate from the large cages to the Sr sites. This ultimately results in an inhibition of the overall level of Ni’ + reduction. In the case of dilute nickel samples, the higher heating rates lower the levels of conversion, Pig. 5. The situation is completely reversed for the higher nickel content samples, as illustrated by NiKY-62.5 and NiNaY-63.1 in Fig. 5. This can be considered as a further manifestation of the dominating influence of the acid function in dictating both the extent of Ni(0) formation and the hydrogenation activity. A lower level of Ni2+ reduction generates a lower level of surface Briinsted acidity [23]. Therefore a fine interplay exists between the mass of supported nickel metal and the number and strength of surface hydroxyls generated during hydrogen reduction which dictates the overall level of benzene conversion. Although the level of reduction of Ni2+ ions in the nickel-rich samples is much lower at the higher heating rates, the reduced associated surface acidity serves to promote overall conversions and prolong the lifetime of the catalysts. This effect is further propagated by calcining the samples prior to reduction, Table 5. Introduction of a precalcination step lowers the extent of Ni2+ reduction at each level of exchange [20] and hence reduces the activity of the dilute nickel samples and increases the activity of the concentrated nickel samples. The present data again serve to illustrate how hydrogenation activity associated with the metal function is intrinsically related to the acid function and the measured conversions represent an overall equilibrium between the dual functions.

102

TABLE 5 Effect of precalcination of a range of nickel-exchanged Y zeolites in a 120 cm3 min-’ stream of nitrogen at 723 K prior to reduction in a 120 cm3 min-’ steam of hydrogen at 723 K for 18 h on the level of benzene hydrogenation (after 6 h on stream) at 473 K; W/F’= 134.1 g mol-’ h Zeolite sample

NiLiY-21.2 NiLiY-43.1 NiLiY-63.7 NiNaY-22.8 NiNaY-35.7 NiNaY-63.1 NiKY-23.5 NiKY-30.5 NiKY-62.5 NiRbNaY-27.3 NiRbNaY-35.1 NiRbNaY-59.1 NiCsNaY-22.0 NiCsNaY-31.1 NiCsNaY-54.8

Effect

Ni’+/U.C.

6.2 12.4 18.5 6.6 10.4 18.3 6.8 10.3 18.1 7.9 10.2 17.1 6.4 9.0 15.9

Cyclohexane (mol%) Reduced at 723 K

hecalcined at 723 K

49.3 3.2 0.5 59.1 6.1 0.9 76.9 39.0 68.4 80.1 86.6 73.0 84.0 89.8 77.9

38.1 36.3 38.9 42.1 18.4 20.5 45.4 42.2 74.4 39.8 54.2 75.2 34.2 50.3 71.2

of surface acid poisoni~ As the presence of surface acidity is the main cause of catalyst deactivation, it was decided to investigate the effect of neutralizing the acid centres by the adsorption of base. Ammonia was chosen because (a) it is sufficiently small to diffuse into the pore structure and ‘poison’ the small cage Briinsted acid sites and (b) it is inert under the stated reaction conditions. It has already been shown that ammonia pretreatment raises the level of Ni2+ cation reduction [ 261. Ammonia-treated reduced zeolites were further contacted at 423 K with ammonia (in a 120 cm3 ruin-’ flowing stream) for 15 min, prior to catalysis. In Fig. 6 the activity profile obtained on carrying out the reaction immediately after adsorption of ammonia is shown. The resultant proille is radically different from those observed in Fig. 1, in that initially the activity rises rapidly and then begins to level out. In addition, as the reaction proceeds the zeolite samples contained progressively less ammonia. These results may be interpreted as arising from an initial blocking of the zeolite pores with ammonia which retards any intracrystalline benzene diffusion; as the reaction proceeds some of the ammonia is dislodged, resulting in the observed increase in activity with time. The inference is that the major fraction of active metal is located within the zeolite pore structure, as the presence of the occluded ammonia molecules restrict passage to the internal active sites and serve to drastically inhibit hydrogenation. The reaction therefore extends to the internal surface, and the external surface area is a negligible function of the total catalytically effective surface area. Flushing with hydrogen at 473 K

103

. ?72 L? P _o 54. x 0 . 36.

+m fm.

A’*

.

.

A

I

.

.

.

.

.

(I

: u

+m

l

:

l

*

l

2 . a

A

.

.

*

.

.

,&

.

=

1

.

76. +.

. 18.

.

.

.

.

.

.

l .+:

1.5 Ttme

30 on stream

/

4.5 hr:

.

x

x

. x

52. .

.

.

XII

a, s

.



+

$64.

2 :

.

E 88.

x

.

A

A

.. 6

12

18 Ni2’/

24

UC.

Fig. 6. Effect of ammonia treatment on time-on-stream benzene hydrogenation activity of NiNaY-48.7 (14.1 Ni2’/U.C.) (T=473 K, W/F=134.1 g mol-’ h): (I) catalysis immediately after ammonia adsorption; (A) after the catalyst was heated in hydrogen for 13 h; (0) on adsorption of ammonia the fresh catalyst was heated in hydrogen for 5 h at 423 K prior to catalysis.

Fig. 7. Variation of benzene hydrogenation (after 6 h on stream) as a function of nickel loading over a range of ammonia-treated ion-exchanged NiLiY (X), NiNaY (A), NiKY (O), NiRbNaY (I) and NiCsNaY (+) samples; T= 473 K, W/F= 134.1 g mol-’ h.

for 5 h removes most of the occluded ammonia, with an accompanying increase in activity. Nevertheless, this steady state conversion is lower than that resulting from the normal pretreatment (also shown in Fig. S), i.e. adsorption of ammonia at 423 K followed by thorough flushing with hydrogen at 423 K for 5 h prior to catalysis at 473 K. It should be noted that this ammonia pretreatment does not affect the initial portion of the activity profile, but greatly modifies the level of conversion after 6 h on stream. MCPa was also absent from the product mixtures, although trace amounts were found in the initial samples collected during catalysis over the nickel-rich NaY and LiY samples. The level of cyclohexane formation after ammonia pretreatment as a function of nickel loading is plotted in Pig. 7. It can be clearly observed that poisoning of surface acidity greatly enhanced the hydrogenation activity of the catalysts; mol% cyclohexane now increases with increasing nickel loading. However, even after flushingthe ammonia-treatedsamples in hydrogen for 4 h at 473 K, there is still a quantity of surplus ammonia within the zeolite pores due to the very high afhnity of the zeolite framework for these small polar molecules. This residual ammonia actually inhibits the catalytic activity of the dilute nickel samples by restricting access to the internaI active sites. In the case of the higher-loaded samples, the promotional effect which results from poisoning of the acid sites far outweighs any diffusional constraints. The effectiveness of pyridine and quinoline in promoting the reaction are presented in Table 6. In agreement with the earlier degree of Ni’ + cation reduction [ZS] and surface acidity [23] studies, the relative efficiencies of the acid poisons increase in the order: quinoline < pyridine
104 TABLE 6 Comparison of the relative efficiencies o f ammonia, pyridine and quinoline as acid poisons in promoting the level of cyclohexane formation (after 6 h on stream) at 473 K over a range of nickel-exchanged Y zeolites; W / F = 134.1 g tool -~ h Zeolite sample

NiLiY-21.2 NiLiY-43.1 NiLiY-63.7 NiNaY-22.8 NiNaY-35.7 NiNaY-63.1 NiNaYo90.1 NiKY-23.5 NLKY-36.6 NiKY-62.5 NiKY-86.6 NiRbNaY- 18.6 NiRbNaY-35.1 NiRbNaY-59.1 NiCsNaY- 16.4 NiCsNaY-31.1 NiCsNaY-54.8 NiNH4Y-20.7 NiNH4Y-52.2 NiNH4Y-69.1

Cyclohexane (mol%) Ammonia treated

Pyridine treated

Quinoline treated

Untreated

55.1 62.2 67.9 51.2 60.1 73.1 89.4 74.0 86.7 94.2 95.3 66.9 89.3 92.2 65.4 90.1 94.3 40.1 43.4 31.9

42.2 44.1 50.0 40.4 48.8 59.9 68.4 48.0 57.5 76.6 76.3 58.2 67.7 77.1 58.8 74.2 75.6 34.2 30.1 24.1

38.4 43.2 48.5 34.3 39.5 46.3 53.9 36.3 44.1 64.4 66.9 48.6 55.3 61.0 50.1 58.3 66.2 26.1 28.3 16.2

49.3 3.2 0.5 59.1 6.1 0.8 7.8 76.9 39.0 68.4 30.3 69.3 86.6 73.0 72.3 89.8 77.9 -

In previous studies [25, 26], it has been shown that nickel-exchanged NH4Y zeolites exhibited such a high level of surface acidity that the reduction of supported Ni 2+ ions was minimal. As a direct consequence, these samples proved inactive towards benzene hydrogenation, Table 6. Contacting the NiNH4Y samples with acid poisons raised the level of Ni 2+ reduction [26] with the introduction of a degree of hydrogenation activity, Table 6. However, this level of activity remained much lower than that exhibited by the alkali metal ion-based zeolites.

Effect of catalyst preparation by impregnation The reduction of nickel Y zeolites prepared by impregnation has been shown to result in the generation of a greater mass of nickel metal compared with samples prepared by ion exchange in the case of NiLiY, NiNaY and NiKY systems [26]. The lower degrees of Ni 2+ reduction (except for the highest nickel-loaded samples) exhibited by the NiRbNaY and NiCsNaY impregnated samples has been explained by the stronger influence of Rb + and Cs + in stabilizing the Ni2+ ions within the crystalline lattice [26]. Reduction of the impregnated samples generates water, NH3, NO2 and NO

105

byproducts which are carried away in the hydrogen stream without the formation of surface hydroxyls, with the result that these samples exhibit minimal surface acidity [23]. In fact the nickel-impregnated NH4Y samples promoted the formation of minor amounts of cyclohexane (m 10% after 6 h on stream). The time-on-streamactivity profiles for the impregnated samples are characterized by rapid attainmentof a steady state conversion, as illustrated in Fig. 1. The decrease in activity over the first hour on stream is much less pronounced for the impregnated samples. This can be considered to result from a lower relative interaction of benzene with the impregnated surface due to overcrowding of the supercages [25]. Activity VS.nickel loading for a range of nickel-impregnated Y zeolites is plotted in Fig. 8. The variation of activity with nickel content mirrors that observed for the degree of Ni2’ cation reduction [26], reinforcing the strong influence of the alkali metal co-cation on the Y zeolite microchemistry.

as

Eflect of coke deposition For a catalyst to be commercially viable, it is essential that it can be used in a number of reaction cycles without an appreciable loss in catalytic activity. In the case of the lower nickel-loaded samples, activity is recoverable to a certain extent (more successfully in the case of the less acidic NiRbNaY and NiCsNaY samples) by flushing in hydrogen atmosphere. This was not possible for the higher-loaded zeolites which exhibited irreversible deactivation, particularly characteristic of the NiLiY samples (Table 7), which showed zero activity after 3 reaction cycles. By comparison, the NiCsNaY samples still exhibited residual hydrogenation activity after 20 reaction cycles. Catalyst deactivation during the hydrogenation of benzene has been shown to result from the generation of cyclohexylbenzene and dicyclohexylbenzene molecules within the zeolite pores which are too large to diffuse out and which prevent passage of benzene molecules to, and cyclohexane molecules from, the internal active sites [ 19 1.The extent of coke formation was obtained in this study by measuring the residual carbon content of the spent catalysts by microanalysis; the results are presented in Fig. 9 and Table 8. It can be

+ 5

10

15

20

NI~+/U.C. 8. Variation of benzene hydrogenation (after 6 h on stream) as a function of nickel loading over a range of impregnated NiLiY m, NiNaY (A), NiKY (O), NiRbNaY (I) and NiCsNaY (+) samples, T-473 K, W/F= 134.1 g mol-’ h.

Fig.

106 TABLE 7 Variation in benzene hydrogenation (after 6 h on stream) at 473 K with reaction cycle over a range of nickel-exchanged Y Zeolites; W/F= 134.1 g mol-’ h Reaction No.

Cyclohexane (mol%)

1 2 3 4 5 6 10 13 16 20

NiCsNaY-44.3

3.2 0.8

87.3 77.6 69.2 56.9 43.8 31.3 18.3 9.6 3.2 -

93.6

x

f u” 54.

B c 1.8.

NiKY-49.1

.

= z2.

2 5 3.6.

NiLiY-43.1

1 x

.

.

87.4 80.4 70.8 58.8 48.5 31.2 20.2 12.2 0.8

l

. l

.

.*

. .*I . AA *AA . . l . n XIA .. . + + A*..** .+ 6 12 16 Ni2’/ U.C. x

24

F’ig. 9. Variation of coke formation during benzene .hydrogenation(after 6 h on stream) as a function of nickel loading over a range of ion-exchanged NiLiY @), NiNaY (A), NiKY (O), NiRbNaY (m) and NiCsNaY (+) samples; T=473 K, W/F= 134.1 g mol-’ h.

seen that the spent samples contain a considerable level of carbon which increases with increasing nickel exchange and increasing zeolite acidity. This is consistent with the earlier mechanistic report [ 19 ] which showed that coke formation was promoted at the acid sites. The observed order of increased coking, i.e. NiCsNaY < NiRbNaY < NiKY < NiNaY < NiiY, is consistent with the order of increasing sample acidity [ 231 and increasing catalyst deactivation. Coke formation is therefore not simply a function of time [ 11, 121 but is dependent on a number of parameters and, in agreement with F’roment and Bishop [ 131, is promoted during the course of the main reaction. Contacting the reduced acidic zeolites with a neutralizing base, which has been shown to raise hydrogenation levels, also inhibits coke formation as shown in Table 8. Ammonia again proves to be the most effective poison; pyridme and quinoline, due to their size, are unable to gain access to the small cages and as a result can only adsorb on the supercage acid sites. One can therefore infer that the small cage acid sites are also active in promoting coke formation.

107 TABLE 8 Comparison of the relative efficiencies of ammonia, pyridine and quinoline as acid poisons in suppressing the level of coke formation during the hydrogenation of benzene (after 6 h on stream) at 473 K over a range of nickel-exchanged Y zeolites; W/F= 134.1 g mol-’ h Zeolite sample

NiLiY-21.2 NiLiY-43.1 NiLiY-63.7 NiNaY-35.7 NiNaY-63.1 NiNaY-90.1 NiKY-36.6 NXY-62.5 NiKY-86.6 NiRbNaY-59.1 NiCsNaY-54.8

Residual carbon (%) Ammonia treated

Pyridine treated

Quinoline treated

Untreated

0.8 1.5 0.3 1.0 2.2 0.3 0.7 1.4 0.5 0.3

0.5 1.3 2.4 0.7 2.5 3.5 0.5 1.8 2.5 1.4 0.8

0.7 1.5 3.1 0.9 2.9 4.0 0.6 2.0 3.2 1.6 0.9

2.0 4.2 6.7 2.7 5.1 7.6 1.0 4.3 6.6 2.0 1.3

The latter is understandablein terms of proton mobility and a time-fluctuating Briinsted acidity in which the protons are continuously moving within the zeolite lattice [ 271. The higher levels of cyclohexane formation resulting from pyridine relative to quinoline treatment is consistent with a higher level of Ni’ f reduction exhibited by the latter [ 261. hkgenmatim of spen.2 [email protected] Treatment of reduced zeolites with ammonia prior to catalysis represents one possible avenue for prolonging the lifetime of the catalysts. However, coke formation increases with each reaction cycle and complete deactivation ultimately occurs. Reclamation of spent samples by oxidative regeneration represents an alternative route. High temperature oxidation of the used zeolite catalysts achieved almost total removal of coke, as evidenced from carbon microanalysis, and in fact resulted in increased catalytic activity, as shown in Fig. 10. Thermogravimetric scans during the oxidative regeneration process revealed additional weight loss (not observed for heat treatment in nitrogen) in the temperature range 673-773 K which must be due to the loss of carbonaceous deposits. The consequences of such a reoxidation/reduction treatment on the size of the nickel crystallites was also considered. Reoxidation of the nickel zeolites in oxygen at 723 K for 2 h followed by further contact with hydrogen at 723 K for 12 h resulted in redispersion of supported nickel metal, with an overall lower crystallite size, Table 9. The resultant fmer dispersion of supported nickel metal translates into a more effective catalytic system with associated higher benzene conversions. This phenomenon is not apparent for the higher nickel-loaded catalysts (> 10 Ni’ + /UC.), as the higher levels of surface BrSnsted acidity cancel the benefits of enhanced metal

108

0

g60.

+ t

t

.

.

n

+ .

+ .

z 560. G A po-

.

t

. + t . .

l

. A x

.

. .

$0.

t

.

I, . .

.

x

x

1

-I-

A 5 3

. x

. x 5

4 1

Regenerat eon

l

.

1

. 3

0

Reactfon

10. Effect of 5 oxidative regeneration cyclesfollowed by 4 reaction cycles on the level of benzene hydrogenation over NibiY-43.1 (12.4 Ni’+/U.C.) (K), NiNaY-35.7 (10.4 Nis’/U.C.) (A), NiKY-35.6 (10.3 Ni”AJ.C.) (O), NiRbNaY-35.1 (10.1 Ni2+AJ.C.) (I) and NiCsNaY-31.0 (9.0 Ni2+/U.C.) (+); T=473 K; W/F=134.1 g mol-’ h. Fig.

TABLE

9

Effect of oxidative Y zeolites Zeolite sample

regeneration

on the size of nickel metal particles supported

Ni2+/U.C.

Nickel crystallite diameter Reduced 723 K

NiLiY-21.2 NiLiY-43.1 NiLiY-63.7 NiNaY-22.8 NiNaY-48.8 NiNaY-63.1 NiKY-23.5 NiKY-49.1 NiKY-62.5 NiRbNaY-47.4 NiRbNaY-59.1 NiCsNaY-44.3 NiCsNaY-54.8

6.2 12.4 18.5 6.6 14.1 18.3 6.8 14.2 18.1 13.7 17.1 12.8 15.9

34.9 44.6 52.1 38.7 55.1 59.0 34.8 50.1 55.6 42.4 46.7 31.1 35.4

at

on a range of

(run)

Reoxidized at 723 K and reduced at 723 K 32.3 41.8 46.7 33.5 48.3 50.3 32.1 46.2 47.3 36.2 38.7 26.9 26.7

dispersion. On resuming, catalysis without further regeneration, the levels of conversion again begin to fall off, as shown in F’ig. 10. Extensive regenerations (up to 20 reaction cycles) expose the zeolites to high temperature water vapour which may lead to irreversible deactivation of the catalysts due to deahnnination. Conclusions

The variation of benzene hydrogenation with nickel loading (or the nature of the alkali metal co-cation) for both ion-exchanged and impregnated samples

109

closely follows the previously observed changes in Ni2+ reduction. At high nickel loadings (> 10 Ni’+/U.C.), the effects of surface acidity are overwhelming; high levels of acidity promote MCPa and coke formation and are responsible for catalyst deactivation. The temperature of maximum benzene hydrogenation is lowered with increasing surface acidity. Acid poisoning by adsorption of ammonia markedly enhances the levels of conversion of the higher nickel-loaded samples. Regeneration of the zeolite catalysts by burning off occluded coke restores and, in the case of dilute nickel samples, raises the catalytic activity. Nevertheless, after prolonged ammonia or oxidative regeneration treatments the catalysts exhibit irreversible decay.

References 1 J. Oudsr, Catal. Rev. Sci. Eng., 22 (1980) 171. 2 C. H. Bartholomew, P. V. Agrswsl and J. R. Katser, in D. D. Eley, H. Pies and P. B. Weiss (eds.), Advances in Catalysis, Academic Press, New York, 1984, p. 135. 3 S. Nsrsysnsn, St&. Su?$ Sci. CataL, 12 (1982) 245. 4 N. P. Davidova, M. L. Vslcheva and D. M. Shopov, Stud. Su?$ Sci. CataL., 5 (1980) 285. 5 G. Schulz-Ekloff and J. K. Blum, Sz& Sci., 183 (1987) 216. 6 B. Coughkm and M. A. Kesne, Zeolites, II (1991) 2. 7 R. Hugues, Deactivation of Catalysts, Academic Press, New York, 1984. 8 J. R. Anderson and M. Boudsrt, Cat&y&, Science and Technology, Springer, Berlin, 1984. 9 A. B. Stiles (ed.), Catalyst Supports and Supported Catalysts, Butterworths, London, 1987. 10 A. Voorhies, Ind. Eng. Chem., 37 (1945) 318. 11 M. R. Viner and B. W. Wojciechowski, Can. J. Chem Eng., 62 (1984) 870. 12 R. P. L. Apsil, J. B. Butt and J. S. Drsnoff, J. CataL, 92 (1985) 187. 13 G. F. Froment and K. B. Bishoff, Chemical Reactor Analysis and Design, Wiley, New York, 1979. 14 R. E. Elkngham and J. Garret, J. AppL In&. Catal., 3 (1984) 25. 15 H. W. Pennline and S. S. Pollack, Ind. Eng. Chem. Prod. Res. Dev., 25 (1986) 11. 16 T. Hano, F. Nsksshio and K. Kusunoki, J. [email protected] Eng., 9 (1976) 209. 17 R. G. Copperthwsite, G. J. Hut&rings, P. Johnston and S. W. Orchard, J. Chem. Sot., Faraday Trans. 1, 82 (1986) 1007; J. Ckm. Sot., Chem. Cwnmun., (1985) 645. 18 B. Coughlsn and M. A. Kesne, Zeolites, 11 (1991) 12. 19 B. Coughlsn and M. A. Kesne, Zeolitas, II (1991) 483. 20 B. Coughlsn and M. A. Kesne, J. Cati., 123 (1990) 364. 21 E. M. Flsnigen, H. Khstsmi and H. A. Szymsuski, Adv. Chem. Ser., IOZ (1971) 201. 22 A. Vogel,. PracticaL Organic Chemistry, Longmsns, London, New York, 1978. 23 B. Coughlsn and M. A. Kesne, J. CoUoid. In.&& Sci., 137 (1990) 483. 24 B. Coughlsn and M. A. Kesne, J. Chem. Sot., Faraday Trans. I, 86 (1990) 3961. 25 M. A. Kesne, Ph. D. Thesis (Vols. I and II), National University of Ireland, 1988. 26 B. CoughIan and M. A. Kesne, J. Cat&., in press. 27 D. Freude, W. Oehme, H. Schmiedel and B. Staudte, J. Cat&., 32 (1974) 343.